人工智能要考虑到溯因推理  

Considering Abduction in Artificial Intelligence

作　　者： 姚洲(译) Randy LaPolla;Yao Zhou(Institute for Advanced Studies in Humanities and Social Sciences,Beijing Normal University at Zhuhai,Zhuhai 519087,Guangdong,China;School of Humanities,Hangzhou City University,Hangzhou 310015,Zhejiang,China)

机构地区：[1]北京师范大学珠海校区人文和社会科学高等研究院语言科学研究中心,广东珠海519087 [2]浙大城市学院人文学院,浙江杭州310015

出　　处：《长江学术》2023年第4期109-113,共5页Yangtze River Academic

摘　　要：ChatGPT可以搜索大量数据,并根据可能组合搭配的统计概率,拼凑出看似人类生成的文本。这并不令人惊讶,因为所使用的数据本身就是人类创造的文本。但是这些算法并不理解它们正在写什么,也没有超越狭义人工智能的范围。要实现通用人工智能,需要这些算法模仿人类的意义创造能力,这取决于对方对交际者交际意图的溯因推理。溯因推理本质上是对交际者为何做了他们所做之事(可能是语言行为或仅仅只是手势或面部表情,或三者兼具)的猜测。它是一种因果推理,提出某现象的所以然,这在创造一般性的知识以及推断他人执行某种行为时的动机时都有应用,包括在交际中的使用。这种因果推理是我们意识的核心。ChatGPT can search a large amount of data and,on the basis of the statistical probability of possible collocations,can put together something that seems like human-produced text,which should not be surprising,as the data used is human text.But these algorithms do not understand what they are writing,and cannot go beyond Artificial Narrow Intelligence.To achieve Artificial General Intelligence will require algorithms to imitate human meaning creation,which depends on abductive inference of the communicative intention of the communicator by the addressee.Abductive inference is essentially guesses as to why the communicator did what they did(which may be languaging or just gesture or facial expression,or all three).It is causal reasoning,asking why some observed phenomenon is the way it is,and this is applied in creating general knowledge,and in inferring the motivations of others when they perform some action,including its use in communication.This causal reasoning is the core of our consciousness.

关 键 词：人工智能 ChatGPT 溯因推理 归纳推理 交际行为 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] B812[自动化与计算机技术—控制科学与工程]